Re: lynx-dev DOS patch for dev.8

[Doug Kaufman]

On Sun, 29 Aug 1999, Klaus Weide wrote:

> On Sun, 29 Aug 1999, Doug Kaufman wrote:
> 
> > I found that telnet was no longer working properly under DOS, probably
> > because of changes from WATT-32. Stopping curses before invoking
> > telnet seems to fix it. This now works both with MSKERMIT and the
> > CUTCP binary, tn3270.exe.
> 
> This shoud not be necessary.  Curses gets turned off in getfile()
> before the HTLoadAbsolute() that ultimately ends up in remote_session().
> So it seems your are turning curses off twice, and then turning it on
> twice.  (Unless DJGPP doesn't go through the common code in getfile(),
> but I didn't see it being diverted.)

It actually doesn't require a full "stop_curses". I just recompiled
with _eth_release() and _eth_init() instead of stop_curses() and
start_curses(). This seems to work also. These are part of the DOS
stop_curses and start_curses functions and should have been handled if
stop_curses was invoked. I'll try to follow the code logic when I get
a chance. Sometimes the interactions of DOS programs, each using its
own TCP stack, trying to share the same PPP connection, are confusing.
I am amazed that these programs with different stacks can coexist as
well as they can.
 
> Or is the start/stop_curses loginc in getfile broken?  Then we should
> fix it for all systems.
> 
> Maybe it makes sense to _move_ the start/stop_curses from getfile() to
> further down in the calling sequence, but I don't see any reason why
> that should be DOS specific.

I don't know enough to comment on the above.
 
> If those __djgpp_set_ctrl_c and _go32_want_ctrl_break always have to be
> done together with start/stop_curses, shouldn't they become part of the
> start/stop_curses functions instead?

I didn't think that they ALWAYS went together. I'll try to see if
there are instances where they don't.
                              Doug
__
Doug Kaufman
Internet: address@hidden (preferred)
          address@hidden